2

如何给MySQL添加自定义语法 ? - 秋风五丈原

 2 years ago
source link: https://www.cnblogs.com/seancheer/p/16551515.html
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
neoserver,ios ssh client

如何给MySQL添加自定义语法 ?

MySQL语法解析器用的bison(即yacc)来实现的,而词法解析是自己来实现的,涉及到的token都在文件lex.h里面,然后通过Lex_input_stream 里面相关的函数,解析client的sql字节流(其中会通过提前构造好的hash表帮助快速找到对应symbol,相关代码在sql_lex_hash.cc里面,转换为token,交给bison进行语法解析。

为了给MySQL添加一个新的语法,我们必须添加新的token(如果有新增),以及增加新的语法(sql_yacc.yy)里面。本文以给create table增加一个新的options为例,来演示如何给MySQL新增一个语法。最终的效果如下:

create table t1 (
  id int primary key,
  name varchar(100)
) global_partition by hash(id) partitions 10; //global_partition by为新增语法,global_partition为新增token

涉及到的修改文件如下:

sql/lex.h //token
sql/parse_tree_nodes.cc
sql/parse_tree_nodes.h
sql/parse_tree_partitions.cc
sql/parse_tree_partitions.h
sql/parser_yystype.h
sql/sql_yacc.yy

2 新增关键词(token)

文件:sql/lex.h

static const SYMBOL symbols[] = {
    /*
     Insert new SQL keywords after that commentary (by alphabetical order):
    */
    //省略部分代码
    {SYM("GLOBAL_PARTITION", GLOBAL_PARTITION_SYM)}, //注意按照字典序进行添加。
    //省略部分代码
};
  

按照上面的格式添加即可

3 新增语法

文件:sql/sql_yacc.yy

该文件为bison的语法,关于bison语法可以查看这里下面凡是注释标有###为新增部分,没有标的注释是为了方便理解

%token<lexer.keyword> GLOBAL_PARTITION_SYM 1309            /* seancheer */  //### 声明上一步添加的token,声明了才可以使用,编号1309选择一个未使用的就行
%type <global_partition_clause> global_partition_clause //### 声明新增加的数据结构,后面会介绍

create_table_stmt:
          CREATE opt_temporary TABLE_SYM opt_if_not_exists table_ident
          '(' table_element_list ')' opt_create_table_options_etc //最后一个标记在YYSTYPE中对应的是create_table_tail, 后面会介绍
          {
            $$= NEW_PTN PT_create_table_stmt(YYMEM_ROOT, $1, $2, $4, $5,
                                             $7,
                                             $9.opt_create_table_options,
                                             $9.opt_partitioning,
                                             $9.opt_global_partitioning, //### 赋值给对应参数,该构造函数需要新增,后面会介绍
                                             $9.on_duplicate,
                                             $9.opt_query_expression);
          }
        | CREATE opt_temporary TABLE_SYM opt_if_not_exists table_ident
          opt_create_table_options_etc
          {
            $$= NEW_PTN PT_create_table_stmt(YYMEM_ROOT, $1, $2, $4, $5,
                                             NULL,
                                             $6.opt_create_table_options,
                                             $6.opt_partitioning,
                                             $6.opt_global_partitioning, //### 赋值给对应参数,该构造函数需要新增,后面会介绍
                                             $6.on_duplicate,
                                             $6.opt_query_expression);
//partition相关的语法                                             
opt_create_partitioning_etc:
          partition_clause opt_duplicate_as_qe //这里是原生的partition表语法
          {
            $$= $2;
            $$.opt_partitioning= $1;
          }
        | global_partition_clause opt_duplicate_as_qe //### 此处是新增的global_partition语法,
          {
            $$= $2;
            $$.opt_global_partitioning= $1;
          }
        | opt_duplicate_as_qe
        ;

//### 下面为重点,新增的global_partition语法,可以看到,用到了新增的token
global_partition_clause:
          GLOBAL_PARTITION_SYM BY part_type_def opt_num_parts
          {
            $$= NEW_PTN PT_global_partition($3, @4, $4);
          }
        ;

4 类似于PT_partition添加对应的数据结构global_partition_clause

文件:parser_yystype.h:该文件是bison(yacc)运行的一环,代替bison内置的YYSTYPE的,当bison对相关语法解析后,需要构造相关的数据结构,通过对YYSTYPE的自定义,就可以实现构造自定义数据结构的目的了。添加我们自定义的数据结构代码如下:

union YYSTYPE {
  PT_sub_partition *opt_sub_part;
  PT_part_type_def *part_type_def;
  PT_partition *partition_clause;
  PT_global_partition *global_partition_clause; //新加数据结构
  
  struct {
    Mem_root_array<PT_create_table_option *> *opt_create_table_options;
    PT_partition *opt_partitioning;
    PT_global_partition *opt_global_partitioning; //同时注意添加到create_table_tail里面,因为create table语法会有该操作
    On_duplicate on_duplicate;
    PT_query_primary *opt_query_expression;
  } create_table_tail;
};
static_assert(sizeof(YYSTYPE) <= 40, "YYSTYPE is too big"); //因为struct里面添加了一个成员变量,所以该union需要的空间也会变大,因此注意修改这一行

下面内容介绍PT_global_partition数据结构,为了保持和MySQL习惯一致,新增加的数据结构放在了

sql/parse_tree_nodes.cc sql/parse_tree_nodes.h sql/parse_tree_partitions.cc sql/parse_tree_partitions.h

四个文件里,理论上可以放在任何地方。可根据自身需求添加对应数据结构:

文件:sql/parse_tree_partitions.h sql/parse_tree_partitions.cc

/**
新增数据结构
*/
class PT_global_partition : public Parse_tree_node {
    typedef Parse_tree_node super;

    PT_part_type_def *const part_type_def;
    const POS part_defs_pos;
    uint num_parts;
public:
    partition_info part_info;
public:
    PT_global_partition(PT_part_type_def *part_type_def, const POS &part_defs_pos,
                        uint opt_num_parts)
            : part_type_def(part_type_def),
              part_defs_pos(part_defs_pos),
              num_parts(opt_num_parts) {}
    bool contextualize(Parse_context *pc) override;
};

//模仿其原生的实现方式即可
bool PT_global_partition::contextualize(Parse_context *pc) {
    if (super::contextualize(pc)) return true;

    Partition_parse_context part_pc(pc->thd, &part_info, false);
    if (part_type_def->contextualize(&part_pc)) return true;

    if (part_info.part_type != partition_type::HASH) {
        //only support hash partition for shard key
        my_error(ER_PARTITIONS_MUST_BE_DEFINED_ERROR, MYF(0), "NOT HASH");
        return true;
    }

    uint count_curr_parts = part_info.partitions.elements;

    if (part_info.num_parts != 0) {
        if (part_info.num_parts != count_curr_parts) {
            error(&part_pc, part_defs_pos,
                  ER_THD(pc->thd, ER_PARTITION_WRONG_NO_PART_ERROR));
            return true;
        }
    } else if (count_curr_parts > 0)
        part_info.num_parts = count_curr_parts;
    return false;
}

文件:sql/parse_tree_nodes.cc sql/parse_tree_nodes.h

接下来修改create table对应的数据结构,将新增的PT_global_partition添加到create table里面

class PT_create_table_stmt final : public PT_table_ddl_stmt_base {
  PT_partition *opt_partitioning;
  PT_global_partition *opt_global_partitioning; //添加成员变量
  PT_create_table_stmt(
    MEM_ROOT *mem_root, PT_hint_list *opt_hints, bool is_temporary,
    bool only_if_not_exists, Table_ident *table_name,
    const Mem_root_array<PT_table_element *> *opt_table_element_list,
    const Mem_root_array<PT_create_table_option *> *opt_create_table_options,
    PT_partition *opt_partitioning,
    PT_global_partition *opt_global_partitioning, On_duplicate on_duplicate,
    PT_query_primary *opt_query_expression)
    : PT_table_ddl_stmt_base(mem_root),
  m_opt_hints(opt_hints),
  is_temporary(is_temporary),
  only_if_not_exists(only_if_not_exists),
  table_name(table_name),
  opt_table_element_list(opt_table_element_list),
  opt_create_table_options(opt_create_table_options),
  opt_partitioning(opt_partitioning),
  opt_global_partitioning(opt_global_partitioning), //添加构造函数,主要是为了增加对PT_global_partition的赋值操作
  on_duplicate(on_duplicate),
  opt_query_expression(opt_query_expression),
  opt_like_clause(nullptr) {}
  
//在其对应的函数中增加相关逻辑,调用对应的初始化函数contextualize
Sql_cmd *PT_create_table_stmt::make_cmd(THD *thd) {
    if (opt_global_partitioning){
        if (opt_global_partitioning->contextualize(&pc)) return nullptr;
        lex->part_info = &opt_global_partitioning->part_info;
    }
}


About Joyk


Aggregate valuable and interesting links.
Joyk means Joy of geeK